Educational Story-Based Game for Capturing the Learner's Personality

This paper presents an educational serious game designed to teach the C programming language while simultaneously capturing players' in-game behavior to model their personality traits. The authors argue that self-report questionnaires, the dominant method for personality assessment, are vulnerable to socially desirable responding and limited self-awareness, and that behavioral data gathered during play offers a viable alternative. The game is story-based: students act as trainees who review short code snippets, accept or reject them, receive immediate feedback, and engage in dialogue with non-player characters whose responses branch according to the player's choices. Personality was modeled using the Five-Factor (OCEAN) framework, focusing on the Extraversion and Openness traits, with the Big Five Inventory serving as ground truth.

The study involved 107 first-year Computer Science students at the University of Macedonia, 72 male and 35 female, who played the game and then completed the questionnaire. Trait scores were categorized as low, balanced, or high, and three classifiers were compared: k-nearest neighbors, decision trees, and Naive Bayes, with PCA used for dimensionality reduction and selection based on F1-score and ROC/AUC. For Extraversion, decision trees achieved 72.7% accuracy, an F1-score of 0.736, and a Cohen's kappa of 0.54, indicating moderate agreement with the questionnaire. For Openness, Naive Bayes reached 81.8% accuracy, an F1-score of 0.811, and a kappa of 0.636, indicating substantial agreement. Openness modeling excluded the low category due to too few participants in that group.

The authors conclude that personality dimensions can be inferred from in-game choices with moderate to good agreement against the questionnaire. They acknowledge limitations, including a modest and single-institution sample, dialogue content not designed with psychological expertise, and the exclusion of low-Openness cases. Proposed future work includes modeling the remaining OCEAN traits, incorporating a third assessment method, and developing adaptive game elements tailored to learner personality.